Intelligent infusion control method and system for nursing nutrient solution

Abstract

The invention relates to the technical field of medical intelligence, in particular to an intelligent infusion control method and system for a nursing nutrient solution. Patient information is collected through multi-modal data, a basic formula is screened from a formula library through an intelligent matching algorithm, adaptive adjustment is conducted in combination with the clinical state of a patient to generate a personalized formula, and pipeline physical parameters and patient physiological parameters are monitored in real time in the infusion process by calculating an initial infusion parameter set. According to the method, infusion parameters are dynamically adjusted based on a comparison result of a monitoring data stream and an expected target, a comprehensive risk score is calculated by adopting a three-dimensional quantitative factor, graded early warning and self-adaptive emergency processing are realized, and after an infusion period is finished, the infusion effect is evaluated in a mode of combining short-term effect quantitative evaluation and long-term effect trend evaluation. And the evaluation result is used for optimizing an intelligent matching algorithm and dynamic adjustment logic.

Description

Technical Field

[0001] This invention relates to the field of medical intelligent technology, and more specifically, to an intelligent infusion control method and system for nursing nutrition solutions. Background Technology

[0002] Nutritional fluid infusion is a crucial means of ensuring patients' nutritional supply, and its control method directly affects treatment outcomes and patient safety. Traditional infusion control is mainly manual, with medical staff determining parameters such as infusion rate and total volume based on experience or basic formulas (such as weight and age estimations), and executing the infusion through mechanical pumps or gravity infusion devices. This requires regular manual monitoring and adjustment, resulting in low efficiency, high subjectivity, and susceptibility to human error.

[0003] With the development of medical technology, automated infusion devices based on preset programs have emerged. These devices allow medical staff to input fixed parameters (such as target flow rate and infusion duration) and achieve precise infusion through the pump-driven system. However, they are essentially still "open-loop control"—they cannot perceive changes in the patient's physiological state (such as blood sugar fluctuations and gastrointestinal reactions) or abnormalities in the infusion tubing (such as blockages and leaks) in real time. They can only execute according to preset instructions and lack dynamic adaptation capabilities.

[0004] The existing technology has the following technical defects, specifically:

[0005] 1. Existing technologies collect fragmented patient data, which relies heavily on manual input and various types of information are stored in a scattered manner. Nutritional solution formulas are mostly generic, and initial infusion parameters are estimated based on manual experience or simple formulas. This makes it difficult for formulas and parameters to be adapted to the individual patient's condition and nutritional needs. Incomplete data and parameter deviations can easily lead to infusion discomfort. Manual operation can also introduce subjective errors.

[0006] 2. Existing infusion technologies are mostly open-loop control, which only executes according to preset instructions, can only monitor a single parameter or even have no real-time monitoring, and the abnormal handling is mostly a fixed response, such as direct shutdown due to overpressure. This makes it impossible to detect problems such as parameter fluctuations and pipeline abnormalities in a timely manner, and emergency response is either excessive or insufficient, which seriously affects the safety and efficiency of infusion.

[0007] 3. Existing technologies lack a comprehensive evaluation system for infusion effects and lack an algorithm self-learning mechanism. The optimization of infusion protocols relies entirely on the experience summarized manually by medical staff, which makes it impossible to scientifically quantify the infusion effect, hinders the technology from achieving autonomous iteration, makes it difficult to continuously improve the long-term infusion effect, and makes it difficult to adapt to ever-changing clinical needs. Summary of the Invention

[0008] The purpose of this invention is to provide an intelligent infusion control method for nursing nutrient solutions to solve the problems mentioned in the background art.

[0009] To achieve the above objectives, the present invention aims to provide an intelligent infusion control method for nursing nutrition solutions, comprising: S1, multimodal patient data acquisition: acquiring multimodal data including basic physiological data of patients, disease diagnosis information and individualized nutritional requirements parameters through connected medical devices and information systems.

[0010] S2. Personalized Nutritional Solution Formula Generation: Based on multimodal data, a basic formula is selected from a pre-stored formula library through an intelligent matching algorithm, and adaptively adjusted according to the patient's specific clinical condition to generate a personalized nutritional solution formula.

[0011] S3. Initialization of infusion parameters: Calculate and determine the initial set of infusion parameters based on the personalized nutrient solution formula and the patient's real-time status.

[0012] S4. Real-time monitoring of the infusion process: During the infusion process, the physical parameters of the infusion tubing and the patient's key physiological parameters are collected in real time by integrated sensors to form a monitoring data stream.

[0013] S5. Dynamic adjustment of infusion parameters: The monitoring data stream is compared with the expected target and safety threshold in real time, and the infusion parameters are dynamically adjusted based on the comparison results.

[0014] S6. Tiered Early Warning and Adaptive Emergency Response: Based on the type and severity of the abnormal events detected, the early warning level is determined, different levels of comprehensive early warning are triggered, and corresponding adaptive emergency operations are executed.

[0015] S7. Infusion effect evaluation and model self-learning: After the infusion cycle is completed, the patient's physiological index changes and infusion data are collected. Based on the patient's physiological index changes and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

[0016] As a further improvement to this technical solution, the basic formula is selected from the pre-stored formula library using an intelligent matching algorithm. The specific implementation method is as follows: Based on the nutrient solution formula library in the database, the proportion information of the core nutrients in each formula is extracted and a nutrient proportion feature vector is constructed. At the same time, based on the collected basic physiological data and nutritional requirement parameters of the patient, the patient's nutritional requirement feature vector is calculated. The cosine similarity algorithm is used to calculate the matching coefficient of the two types of vectors corresponding to the patient for each formula. A preset matching coefficient threshold is extracted from the database. The matching coefficient of the two types of vectors corresponding to the patient for each formula is compared with the preset matching coefficient threshold extracted from the database. Formulas with matching coefficients higher than the preset matching coefficient threshold are selected as candidate formulas. After sorting the formulas from high to low according to the matching coefficient, the formulas whose clinical application frequency meets the preset requirements are selected as the basic nutrient solution formulas.

[0017] As a further improvement to this technical solution, the specific method for generating personalized nutrient solution formula is as follows: After determining the basic nutrient solution formula through an intelligent matching algorithm, based on the collected data related to the patient's specific clinical condition, the core direction and constraints of formula adjustment are clarified. In view of the nutritional metabolism characteristics, organ function status and pathophysiological needs corresponding to the patient's clinical condition, the types, proportions, distribution and content range of core nutrients in the basic formula are adaptively adjusted. Combined with the patient's real-time physiological monitoring data, the adjustment parameters are dynamically calibrated and adjusted to finally generate a personalized nutrient solution formula.

[0018] As a further improvement to this technical solution, the calculation of the initial infusion parameter set is specifically implemented as follows: A personalized nutrient solution formula is extracted; based on the total amount of nutrients in the personalized nutrient solution formula and the patient's basic physiological data, combined with the parameter benchmarks in the clinical nutrition infusion guidelines, the total infusion volume for a single infusion is calculated using a preset algorithm; the infusion interval is determined based on the patient's gastrointestinal tolerance, the urgency of nutritional needs, and the formula viscosity; the initial infusion flow rate is calculated based on the total infusion volume for a single infusion and the infusion interval, combined with the time required for pre-infusion pretreatment; and the above parameters are integrated to form the initial infusion parameter set.

[0019] As a further improvement to this technical solution, the specific implementation method of dynamically adjusting the infusion parameters is as follows: during the infusion process, the physical parameters of the infusion tubing and the patient's physiological parameters are collected in real time to form a monitoring data stream. The monitoring data is compared with the preset expected target value and safety threshold, and the deviation value is calculated. According to the type and degree of deviation, the corresponding parameter adjustment rules are called to determine the adjustment direction and amplitude. The adjustment operation is executed through the drive control module of the infusion pump, and the adjusted monitoring data is continuously collected for closed-loop feedback until the deviation returns to the preset allowable range.

[0020] As a further improvement to this technical solution, the specific method for determining the warning level is as follows:

[0021] Based on a pre-defined abnormal event risk classification library, different types of deviations are assigned corresponding risk weights. Using "deviation type weight," "deviation degree level," and "abnormal duration" as three-dimensional quantification factors, the deviation degree level is converted into a corresponding quantified score, and the abnormal duration is converted into a time quantified value according to a pre-defined interval. The deviation type weight is multiplied by the deviation degree quantified score and the abnormal duration quantified value to obtain a comprehensive risk score. A comprehensive risk score interval corresponding to multiple pre-defined warning levels is set. The calculated comprehensive risk score is mapped to the corresponding interval to determine the final warning level, which is a low level, a medium level, and a high level.

[0022] As a further improvement to this technical solution, the adaptive emergency operation is specifically implemented as follows:

[0023] Based on the warning level and the type of abnormal event, the system calls the pre-stored emergency operation rule library and matches the corresponding adaptive emergency response plan. In the case of a low-level warning, the infusion parameters are slightly adjusted and continuously monitored. In the case of a medium-level warning, the relevant operations are suspended and the infusion mode is adjusted. In the case of a high-level warning, the infusion is immediately terminated and enhanced monitoring of the patient's vital signs is initiated. After the emergency operation is executed, feedback data is collected in real time to evaluate the treatment effect. If the abnormality is not resolved, the warning level is upgraded and a higher-level emergency operation is triggered until the abnormality is under control.

[0024] As a further improvement to this technical solution, the specific method for evaluating the infusion effect is as follows:

[0025] S7.1 Quantitative evaluation of short-term effects: After the infusion cycle ends, calculate the fluctuation index and target achievement rate of key physiological indicators.

[0026] S7.2 Long-term effect trend assessment: After the completion of multiple consecutive infusion cycles, track and calculate the changing trends of the patient's core nutritional status indicators.

[0027] S7.3 Comprehensive Effect Judgment and Data Labeling: Combining the short-term effect quantitative evaluation results and the long-term effect trend evaluation results, the comprehensive effect of this infusion plan is graded according to preset rules; and the judgment result, the corresponding infusion parameters and patient data are jointly labeled to form a labeled training sample, which is stored in the model optimization database.

[0028] A second aspect of the present invention provides a system for intelligent infusion control of nursing nutrition solutions, comprising: a multimodal patient data acquisition module: acquiring multimodal data including basic physiological data of patients, disease diagnosis information and individualized nutritional requirements parameters through connected medical equipment and information systems.

[0029] Personalized nutrient solution formula generation module: Based on multimodal data, it uses an intelligent matching algorithm to select a basic formula from a pre-stored formula library and makes adaptive adjustments according to the patient's specific clinical condition to generate a personalized nutrient solution formula.

[0030] Infusion parameter initialization module: Calculates and determines the initial infusion parameter set based on the personalized nutrient solution formula and the patient's real-time status.

[0031] Real-time monitoring module for infusion process: During the infusion process, the module integrates sensors to collect physical parameters of the infusion tubing and key physiological parameters of the patient in real time, forming a monitoring data stream.

[0032] Dynamic infusion parameter adjustment module: It compares the monitoring data stream with the expected target and safety threshold in real time, and dynamically adjusts the infusion parameters based on the comparison results.

[0033] Tiered early warning and adaptive emergency response module: Based on the type and severity of the abnormal events detected, the module determines the early warning level, triggers comprehensive early warnings of different levels, and executes adaptive emergency operations that match the level.

[0034] Infusion effect evaluation and model self-learning module: After the infusion cycle is completed, the changes in the patient's physiological indicators and infusion data are collected. Based on the changes in the patient's physiological indicators and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

[0035] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0036] 1. This invention centrally collects multimodal patient data through medical equipment and information systems, intelligently selects basic formulas from the formula library using a cosine similarity algorithm, and then adjusts the nutritional components in combination with the patient's specific clinical condition. At the same time, it integrates multiple factors such as the total amount of nutrients in the formula and the patient's gastrointestinal tolerance to calculate the initial infusion parameters, thereby achieving personalized adaptation of formulas and parameters from the source, greatly reducing the errors caused by manual intervention, and effectively reducing the incidence of infusion discomfort problems.

[0037] 2. This invention integrates sensors to collect physical parameters of the infusion tubing and key physiological parameters of the patient in real time. After comparing the deviation with the preset target value, the parameters are dynamically adjusted and feedback is provided in a closed loop. It also calculates a comprehensive risk score based on three-dimensional quantitative factors to classify early warning levels and matches differentiated emergency operations such as fine-tuning, adjustment mode, and termination of infusion. This can not only correct infusion deviations in a timely manner, but also accurately respond to different risks and ensure the stability and safety of the infusion process.

[0038] 3. This invention constructs a complete system that combines short-term physiological indicator quantitative assessment with long-term nutritional status trend assessment. At the same time, the assessment results, patient data, and infusion parameters are labeled as training samples to optimize the intelligent matching algorithm and dynamic adjustment logic, forming a closed-loop iterative mechanism of "infusion-assessment-optimization". Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0041] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example: Please refer to Figure 1 As shown, an intelligent infusion control method for nursing nutrition solutions is provided, including: S1, multimodal patient data acquisition: through connected medical devices and information systems, multimodal data including basic physiological data of patients, disease diagnosis information and individualized nutritional requirements parameters are collected.

[0044] S2. Personalized Nutritional Solution Formula Generation: Based on multimodal data, a basic formula is selected from a pre-stored formula library through an intelligent matching algorithm, and adaptively adjusted according to the patient's specific clinical condition to generate a personalized nutritional solution formula.

[0045] In one specific embodiment, the basic formula is selected from the pre-stored formula library using an intelligent matching algorithm. The specific implementation method is as follows: Based on the nutrient solution formula library in the database, the proportion information of the core nutrients in each formula is extracted and a nutrient proportion feature vector is constructed. At the same time, based on the collected basic physiological data and nutritional requirement parameters of the patient, the patient's nutritional requirement feature vector is calculated. The cosine similarity algorithm is used to calculate the matching coefficient of the two types of vectors of each formula corresponding to the patient. A preset matching coefficient threshold is extracted from the database. The matching coefficient of the two types of vectors of each formula corresponding to the patient is compared with the preset matching coefficient threshold extracted from the database. Formulas with matching coefficients higher than the preset matching coefficient threshold are selected as candidate formulas. After sorting the formulas from high to low matching coefficients, the formulas whose clinical application frequency meets the preset requirements are selected as the basic nutrient solution formulas.

[0046] The specific calculation process of the patient's nutritional needs feature vector is as follows: First, clarify the vector dimension and select core clinical nutrients such as protein, carbohydrates, fat, vitamins (such as B vitamins), and minerals (such as potassium and sodium) as fixed dimensions to ensure that they correspond one-to-one with the dimensions of the nutrient proportion feature vector in the nutrient solution formula library.

[0047] Based on the patient's basic physiological data (weight, age, basal metabolic rate, etc.) and combined with clinical nutrition formulas, the daily requirements of each core nutrient are calculated. For example, the daily protein requirement (g / day) and the requirement corresponding to the proportion of energy provided by carbohydrates are calculated based on weight and metabolic level. Quantitative corrections are made for the patient's special nutritional requirements (such as the need to reduce the proportion of carbohydrates for diabetic patients and the need to restrict the total amount of protein for kidney disease patients).

[0048] All quantified component requirements are normalized to eliminate differences in the dimensions of different components. Finally, the normalized values ​​are arranged in a fixed order of components to form a characteristic vector of the patient's nutritional needs.

[0049] The database has preset matching coefficient thresholds, which are set by professionals. For example, the thresholds are determined based on statistical data of successful clinical nutrition adaptation cases, and the median of the vector matching coefficients between formulas with good adaptation effects in history and their corresponding patients is taken as the initial threshold. At the same time, the thresholds are dynamically adjusted according to the criticality of the core nutritional components. If the adaptation of components involving life safety has a high priority, the threshold can be appropriately increased to improve the matching accuracy.

[0050] The preset requirements for clinical application frequency are set by professionals. For example, the preset requirements for clinical application frequency are based on the clinical safety and effectiveness of the formula, setting a minimum threshold for the number of applications (e.g., ≥50 times) and limiting the upper limit of the incidence of adverse reactions during the application period (e.g., ≤2%). For formulas newly added to the formula library, the frequency requirements can be temporarily reduced and then adjusted to the standard preset requirements after sufficient clinical data has been accumulated.

[0051] When there are multiple candidate formulations and multiple formulations whose clinical application frequency meets the preset requirements, the final basic nutrient solution formulation shall be determined according to the following rules:

[0052] First, extract the historical clinical efficacy data corresponding to this type of formula. The historical clinical efficacy data includes the short-term efficacy pass rate and long-term efficacy pass rate of the corresponding patients.

[0053] Calculate the overall effect score for each formula. Overall effect score = short-term effect pass rate × preset weight + long-term effect pass rate × (1 - preset weight).

[0054] The formulas with the highest overall performance scores, ranked from highest to lowest, were selected as the basic nutrient solution formulas.

[0055] If formulas with the same overall efficacy score exist, further compare the compatibility of the formula with the patient's specific condition, and give priority to formulas with clear clinical application records for the patient's specific condition (such as diabetes or kidney disease).

[0056] In one specific embodiment, the method for generating a personalized nutrient solution formula is as follows: After determining the basic nutrient solution formula through an intelligent matching algorithm, based on the collected data related to the patient's specific clinical condition, the core direction and constraints of formula adjustment are clarified. In view of the nutritional metabolism characteristics, organ function status and pathophysiological needs corresponding to the patient's clinical condition, the types, proportions, distribution and content range of the core nutrients in the basic formula are adaptively adjusted. Combined with the patient's real-time physiological monitoring data, the adjustment parameters are dynamically calibrated to ensure that the adjusted formula is accurately matched with the patient's individual clinical condition, and finally a personalized nutrient solution formula is generated.

[0057] Determining the core direction and constraints of formula adjustment: Based on the patient's specific clinical status data (such as disease diagnosis type, organ function indicators, and metabolic abnormality parameters), the pre-stored clinical nutrition adjustment rule library is called to match the corresponding nutritional component adjustment direction (such as restricting / increasing a certain type of component, or replacing a specific type of component), and the safe threshold range of nutritional components under this clinical status (such as the maximum / minimum allowable content of a certain type of component) is extracted as a constraint.

[0058] Implementation of adaptive adjustments: Based on the determined core direction, the types and combinations of core nutrients in the basic formula are added or deleted (such as adding specific functional nutrients or removing prohibited ingredients), the proportion distribution of each component is adjusted according to the preset ratio range, and the content range is scaled up or down according to the constraints to ensure that the adjusted component combination matches the nutritional needs of the patient's clinical condition.

[0059] Real-time physiological data dynamic calibration: The patient's real-time physiological monitoring data (such as blood glucose, electrolytes, liver and kidney function indicators) are compared with preset target values ​​to calculate the deviation value; based on the magnitude of the deviation value, the corresponding calibration coefficient table is called to perform secondary correction on the component proportion or content of the adjusted formula (the larger the deviation, the larger the correction magnitude is according to the preset gradient) until the real-time monitoring data returns to the target value range, and the final calibration of the personalized formula is completed.

[0060] S3. Initialization of infusion parameters: Calculate and determine the initial set of infusion parameters based on the personalized nutrient solution formula and the patient's real-time status.

[0061] In one specific embodiment, the calculation to determine the initial infusion parameter set is implemented as follows: A personalized nutrient solution formula is extracted; based on the total amount of nutrients in the personalized nutrient solution formula and the patient's basic physiological data, combined with parameter benchmarks in clinical nutrition infusion guidelines, the total infusion volume for a single infusion is calculated using a preset algorithm; the infusion interval is determined based on the patient's gastrointestinal tolerance, the urgency of nutritional needs, and the formula viscosity; the initial infusion flow rate is calculated based on the total infusion volume and the infusion interval, combined with the time required for pre-infusion pretreatment; these parameters are integrated to form the initial infusion parameter set. Furthermore, safety parameter thresholds (such as maximum flow rate limits and minimum interval times) corresponding to the patient's clinical status are used for verification and correction to ensure the safety and applicability of the parameter set.

[0062] Calculation of total infusion volume per session: Based on the total amount of core nutrients in the personalized nutritional solution formula, combined with the patient's basic physiological data (weight, age, basal metabolic rate), a preset algorithm is called (this algorithm integrates the basal metabolic requirement calculation formula with the parameter benchmarks in the clinical nutrition infusion guidelines) to convert the total amount of nutrients into the corresponding infusion volume; at the same time, the calculation results are corrected by referring to the degree of nutritional deficiency of the patient (such as the proportion of weight loss, serum protein level) to obtain the total infusion volume per session.

[0063] Determination of infusion interval: A multi-factor assessment model was constructed, taking the patient's gastrointestinal tolerance (based on previous infusion tolerance records and gastrointestinal function scores), the urgency of nutritional needs (based on the severity of the disease and the priority of nutritional support) and the viscosity of the formula (based on the physical properties of the formula) as input variables. The comprehensive assessment value was calculated by pre-set weight allocation, and the final infusion interval was determined by mapping the assessment value to the pre-set interval range.

[0064] Initial infusion rate calculation: Based on the total infusion volume, after deducting the time required for pre-infusion pretreatment, the remaining time is taken as the effective infusion time. The initial infusion rate is calculated by the ratio of the total volume to the effective infusion time. At the same time, it is verified in conjunction with the patient's corresponding flow rate safety threshold (based on age and underlying disease type). If the calculated value exceeds the threshold, it is adjusted according to the upper limit of the threshold to ensure that the flow rate is within the safe range.

[0065] S4. Real-time monitoring of the infusion process: During the infusion process, the physical parameters of the infusion tubing and the patient's key physiological parameters are collected in real time by integrated sensors to form a monitoring data stream.

[0066] S5. Dynamic Adjustment of Infusion Parameters: The monitoring data stream is compared in real time with the expected targets and safety thresholds, and the infusion parameters are dynamically adjusted based on the comparison results. The adjustment logic includes adjusting the proportion of corresponding components in the nutrient solution based on changes in the patient's real-time physiological parameters.

[0067] In one specific embodiment, the dynamic adjustment of infusion parameters is implemented as follows: During the infusion process, physical parameters of the infusion tubing (such as actual flow rate and tubing pressure) and patient physiological parameters (such as blood glucose and heart rate) are collected in real time to form a monitoring data stream. The monitoring data is compared with preset expected target values ​​(such as target flow rate and target blood glucose range) and safety thresholds (such as pressure upper limit and physiological indicator warning range) to calculate the deviation value. According to the deviation type (physical parameter deviation or physiological parameter deviation) and the degree of deviation (slight, moderate, significant), the corresponding parameter adjustment rules (such as flow rate fine-tuning, infusion pause, and component ratio correction) are called to determine the adjustment direction and amplitude. The adjustment operation is executed through the drive control module of the infusion pump, and the monitoring data after adjustment is continuously collected for closed-loop feedback until the deviation returns to the preset allowable range, thereby realizing the dynamic adaptation of infusion parameters.

[0068] Deviation type determination: Based on the acquisition source and parameter attributes of the monitoring data, the data is classified. Specifically, if the actual flow rate, pipeline pressure, and other data obtained from the physical sensors of the infusion pipeline differ from the corresponding expected target value or safety threshold, it is determined as a physical parameter deviation. If the blood glucose, heart rate, blood oxygen saturation, and other data obtained from the patient's surface physiological monitoring module differ from the corresponding expected target value or safety threshold, it is determined as a physiological parameter deviation. Through a preset parameter attribute tag library, the classification of the monitoring data is automatically identified to complete the rapid determination of the deviation type.

[0069] Deviation degree analysis: For each type of deviation identified, the deviation value (absolute or relative) is first calculated. For physical parameter deviations, the absolute value of "|actual value - expected target value|" is used. For physiological parameter deviations, the absolute value (e.g., heart rate deviation) or relative value (e.g., the proportion of blood glucose deviation to the target range) is selected based on the parameter type. Then, the preset allowable deviation range for each parameter is called, and the calculated deviation value is compared with the allowable deviation range to classify the degree of deviation: if the deviation value does not exceed 1 / 3 of the allowable deviation range, it is judged as a slight deviation; if the deviation value is between 1 / 3 and 2 / 3 of the allowable deviation range, it is judged as a moderate deviation; if the deviation value exceeds 2 / 3 of the allowable deviation range or exceeds the safety threshold, it is judged as a significant deviation. The results of the deviation degree classification are synchronously linked to the parameter adjustment rule base to provide a basis for determining the direction and magnitude of subsequent adjustments.

[0070] S6. Tiered Early Warning and Adaptive Emergency Response: Based on the type and severity of the abnormal events detected, the early warning level is determined, different levels of comprehensive early warning are triggered, and corresponding adaptive emergency operations are executed.

[0071] For each identified warning level, corresponding warning transmission channels and presentation formats are preset; a corresponding comprehensive warning mechanism is activated according to the warning level. Low-level warnings provide prompt feedback through local devices, medium-level warnings send notifications to the mobile terminals of responsible medical staff simultaneously, and high-level warnings adopt a multi-channel linkage approach involving audible and visual alarms, simultaneous push notifications from multiple terminals, and cloud platform filing; the warning information includes patient identification, abnormal event type, abnormal data, warning level, and preliminary handling suggestions to ensure that medical staff can quickly obtain key information.

[0072] In one specific embodiment, the method for determining the warning level is as follows: Based on a preset abnormal event risk classification library, different types of deviations are assigned corresponding risk weights. Among them, deviations of physiological parameters related to the patient's vital signs have a higher weight than deviations of physical parameters of the infusion tubing. Deviations of key parameters that directly affect infusion safety (such as tubing pressure and blood glucose) have a higher weight than deviations of non-key parameters (such as minor fluctuations in infusion flow rate). Using "deviation type weight," "deviation degree level," and "abnormal duration" as three-dimensional quantification factors, the deviation degree level is converted into a corresponding quantification score, and the abnormal duration is converted into a time quantification value according to a preset interval. The deviation type weight is multiplied by the deviation degree quantification score and the abnormal duration quantification value to obtain a comprehensive risk score. A comprehensive risk score interval corresponding to multiple warning levels is preset. The calculated comprehensive risk score is mapped to the corresponding interval to determine the final warning level, which is a low level, a medium level, or a high level.

[0073] Among them, a comprehensive risk score below the first preset threshold is a low-level warning, a score between the first and second preset thresholds is a medium-level warning, and a score above the second preset threshold is a high-level warning. If multiple abnormal events are superimposed, the level is determined according to the highest risk score after superposition, or the warning level is raised by one level according to the preset superposition rules.

[0074] The first and second preset thresholds are determined based on the comprehensive risk scores of historical abnormal events and the corresponding clinical treatment results. By analyzing the score distribution of past low-risk, medium-risk, and high-risk events, the score cutoff point for low-to-medium risk is set as the first preset threshold, and the score cutoff point for medium-to-high risk is set as the second preset threshold. At the same time, the thresholds are dynamically fine-tuned in combination with the risk weight differences of deviation types to ensure that the thresholds can accurately match the actual safety risk level corresponding to different warning levels, avoiding over-warning or under-warning.

[0075] In one specific embodiment, the adaptive emergency operation is implemented as follows: based on the warning level and the type of abnormal event, a pre-stored emergency operation rule library is invoked to match the corresponding adaptive emergency response plan. When the warning level is low, the infusion parameters are slightly adjusted and continuously monitored. When the warning level is medium, the relevant operation is suspended and the infusion mode is adjusted. When the warning level is high, the infusion is immediately terminated and enhanced monitoring of the patient's vital signs is initiated. After the emergency operation is executed, feedback data is collected in real time to evaluate the treatment effect. If the abnormality is not resolved, the warning level is upgraded and a higher-level emergency operation is triggered until the abnormality is under control.

[0076] S7. Infusion effect evaluation and model self-learning: After the infusion cycle is completed, the patient's physiological index changes and infusion data are collected. Based on the patient's physiological index changes and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

[0077] In one specific embodiment, the evaluation of infusion effect is specifically implemented as follows: S7.1, short-term effect quantitative evaluation: after the infusion cycle is completed, the fluctuation index and target achievement rate of key physiological indicators are calculated; the fluctuation index is obtained by calculating the standard deviation or target range of a specific physiological indicator (such as blood glucose) during the infusion period as a percentage of time; the target achievement rate is obtained by statistically analyzing the proportion of patients' physiological indicators falling into the clinical ideal range after the infusion is completed.

[0078] S7.2 Long-term effect trend assessment: After the completion of multiple consecutive infusion cycles, track and calculate the changing trends of the patient's core nutritional status indicators; the core indicators include serum albumin, prealbumin levels and weight change rate; generate a long-term effect score by comparing with the preset clinical improvement target.

[0079] S7.3 Comprehensive Effect Judgment and Data Labeling: Combining the short-term effect quantitative evaluation results and the long-term effect trend evaluation results, the comprehensive effect of this infusion plan is graded according to preset rules; and the judgment result, the corresponding infusion parameters and patient data are jointly labeled to form a labeled training sample, which is stored in the model optimization database.

[0080] See Figure 2 As shown, a system for intelligent infusion control of nursing nutrition solutions is provided, comprising: a multimodal patient data acquisition module: which collects multimodal data including basic physiological data of patients, disease diagnosis information and individualized nutritional requirements parameters through connected medical devices and information systems.

[0081] Personalized nutrient solution formula generation module: Based on multimodal data, it uses an intelligent matching algorithm to select a basic formula from a pre-stored formula library and makes adaptive adjustments according to the patient's specific clinical condition to generate a personalized nutrient solution formula.

[0082] Infusion parameter initialization module: Calculates and determines the initial infusion parameter set based on the personalized nutrient solution formula and the patient's real-time status.

[0083] Real-time monitoring module for infusion process: During the infusion process, the module integrates sensors to collect physical parameters of the infusion tubing and key physiological parameters of the patient in real time, forming a monitoring data stream.

[0084] The infusion parameter dynamic adjustment module compares the monitoring data stream with the expected target and safety threshold in real time, and dynamically adjusts the infusion parameters based on the comparison results. The adjustment logic includes adjusting the infusion ratio of the corresponding components in the nutrient solution in reverse according to the changes in the patient's real-time physiological parameters.

[0085] Tiered early warning and adaptive emergency response module: Based on the type and severity of the abnormal events detected, the module determines the early warning level, triggers comprehensive early warnings of different levels, and executes adaptive emergency operations that match the level.

[0086] Infusion effect evaluation and model self-learning module: After the infusion cycle is completed, the changes in the patient's physiological indicators and infusion data are collected. Based on the changes in the patient's physiological indicators and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A method for intelligent infusion control of nursing nutrient solution, characterized in that, include: S1. Multimodal patient data collection: Collect multimodal data, including basic physiological data, disease diagnosis information and individualized nutritional requirements parameters, through connected medical devices and information systems. S2. Personalized nutrient solution formula generation: Based on multimodal data, a basic formula is selected from the pre-stored formula library through an intelligent matching algorithm, and adaptively adjusted according to the patient's specific clinical condition to generate a personalized nutrient solution formula. S3. Initialization of infusion parameters: Calculate and determine the initial set of infusion parameters based on the personalized nutrient solution formula and the patient's real-time status. S4. Real-time monitoring of the infusion process: During the infusion process, the physical parameters of the infusion tubing and the patient's key physiological parameters are collected in real time through integrated sensors to form a monitoring data stream; S5. Dynamic adjustment of infusion parameters: The monitoring data stream is compared with the expected target and safety threshold in real time, and the infusion parameters are dynamically adjusted based on the comparison results; S6. Tiered Early Warning and Adaptive Emergency Response: Based on the type and severity of the abnormal events detected, the early warning level is determined, different levels of comprehensive early warning are triggered, and corresponding adaptive emergency operations are executed. S7. Infusion effect evaluation and model self-learning: After the infusion cycle is completed, the patient's physiological index changes and infusion data are collected. Based on the patient's physiological index changes and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

2. The intelligent infusion control method for nursing nutrient solution according to claim 1, characterized in that, The method for selecting basic formulas from a pre-stored formula library using an intelligent matching algorithm is as follows: Based on a database of nutrient solution formulas, the proportion of core nutrients in each formula is extracted and a nutrient proportion feature vector is constructed. Simultaneously, based on the collected basic physiological data and nutritional requirement parameters of patients, a patient nutritional requirement feature vector is calculated. The cosine similarity algorithm is used to calculate the matching coefficient between the two types of vectors for each formula corresponding to the patient. A preset matching coefficient threshold is extracted from the database. The matching coefficients of the two types of vectors for each formula corresponding to the patient are compared with the preset matching coefficient threshold extracted from the database. Formulas with matching coefficients higher than the preset matching coefficient threshold are selected as candidate formulas. After sorting the formulas from high to low matching coefficients, the formulas whose clinical application frequency meets the preset requirements are selected as the basic nutrient solution formulas.

3. The intelligent infusion control method for nursing nutrient solution according to claim 2, characterized in that, The specific method for generating the personalized nutrient solution formula is as follows: After determining the basic nutrient solution formula through intelligent matching algorithm, based on the collected data related to the patient's specific clinical condition, the core direction and constraints of formula adjustment are clarified. In view of the nutritional metabolism characteristics, organ function status and pathophysiological needs corresponding to the patient's clinical condition, the types, proportions and content ranges of core nutrients in the basic formula are adaptively adjusted. Combined with the patient's real-time physiological monitoring data, the adjustment parameters are dynamically calibrated and adjusted, and finally a personalized nutrient solution formula is generated.

4. The intelligent infusion control method for nursing nutrient solution according to claim 3, characterized in that, The calculation determines the initial infusion parameter set, and the specific implementation method is as follows: Personalized nutrient solution formulas are extracted, and based on the total amount of nutrients in the personalized nutrient solution formulas and the patient's basic physiological data, combined with the parameter benchmarks in the clinical nutrition infusion guidelines, the total amount of a single infusion is calculated through a preset algorithm. The infusion interval is determined based on the patient's gastrointestinal tolerance, the urgency of nutritional needs, and the viscosity of the formulation. Based on the total infusion volume and infusion interval, combined with the time required for pre-infusion pretreatment, the initial infusion flow rate is calculated; the above parameters are then integrated to form an initial infusion parameter set.

5. The intelligent infusion control method for nursing nutrient solution according to claim 4, characterized in that, The specific method for dynamically adjusting the infusion parameters is as follows: During the infusion process, the physical parameters of the infusion tubing and the patient's physiological parameters are collected in real time to form a monitoring data stream. The monitoring data is compared with the preset expected target value and safety threshold, and the deviation value is calculated. According to the type and degree of deviation, the corresponding parameter adjustment rules are called to determine the adjustment direction and amplitude. The adjustment operation is executed through the drive control module of the infusion pump, and the monitoring data after adjustment is continuously collected for closed-loop feedback until the deviation returns to the preset allowable range.

6. The intelligent infusion control method for nursing nutrient solution according to claim 5, characterized in that, The specific method for determining the warning level is as follows: Based on a pre-defined abnormal event risk classification library, different types of deviations are assigned corresponding risk weights. Using "deviation type weight," "deviation severity level," and "abnormal duration" as three-dimensional quantification factors, the deviation severity level is converted into a corresponding quantified score, and the abnormal duration is converted into a time quantified value according to a pre-defined interval. The deviation type weight is multiplied by the deviation severity quantified score and the abnormal duration quantified value to obtain a comprehensive risk score. A comprehensive risk score interval corresponding to multiple pre-defined warning levels is set. The calculated comprehensive risk score is mapped to the corresponding interval to determine the final warning level, which is a low level, a medium level, or a high level.

7. The intelligent infusion control method for nursing nutrient solution according to claim 6, characterized in that, The adaptive emergency operation is specifically implemented as follows: Based on the warning level and the type of abnormal event, the system calls the pre-stored emergency operation rule library and matches the corresponding adaptive emergency response plan. In the case of a low-level warning, the infusion parameters are slightly adjusted and continuously monitored. In the case of a medium-level warning, the relevant operations are suspended and the infusion mode is adjusted. In the case of a high-level warning, the infusion is immediately terminated and enhanced monitoring of the patient's vital signs is initiated. After the emergency operation is executed, feedback data is collected in real time to evaluate the treatment effect. If the abnormality is not resolved, the warning level is upgraded and a higher-level emergency operation is triggered until the abnormality is under control.

8. The intelligent infusion control method for nursing nutrient solution according to claim 7, characterized in that, The specific method for evaluating the infusion effect is as follows: S7.1 Quantitative evaluation of short-term effects: After the infusion cycle ends, calculate the fluctuation index and target achievement rate of key physiological indicators; S7.2 Long-term effect trend assessment: After the completion of multiple consecutive infusion cycles, track and calculate the changing trends of the patient's core nutritional status indicators; S7.3 Comprehensive Effect Judgment and Data Labeling: Combining the short-term effect quantitative evaluation results and the long-term effect trend evaluation results, the comprehensive effect of this infusion plan is graded according to preset rules; and the judgment result, the corresponding infusion parameters and patient data are jointly labeled to form a labeled training sample, which is stored in the model optimization database.

9. A system for the intelligent infusion control method of nursing nutrient solution according to any one of claims 1-8, characterized in that, include: Multimodal patient data acquisition module: Through connected medical devices and information systems, it collects multimodal data including patients' basic physiological data, disease diagnosis information, and individualized nutritional requirements parameters; Personalized nutrient solution formula generation module: Based on multimodal data, a basic formula is selected from a pre-stored formula library through an intelligent matching algorithm, and adaptively adjusted according to the patient's specific clinical condition to generate a personalized nutrient solution formula. Infusion parameter initialization module: Calculates and determines the initial infusion parameter set based on the personalized nutrient solution formula and the patient's real-time status; Real-time monitoring module for infusion process: During the infusion process, integrated sensors collect physical parameters of the infusion tubing and key physiological parameters of the patient in real time, forming a monitoring data stream; Dynamic infusion parameter adjustment module: compares the monitoring data stream with the expected target and safety threshold in real time, and dynamically adjusts the infusion parameters based on the comparison results; Tiered early warning and adaptive emergency response module: Based on the type and severity of the abnormal events detected, the module determines the early warning level, triggers comprehensive early warnings of different levels, and executes adaptive emergency operations that match the level. Infusion effect evaluation and model self-learning module: After the infusion cycle is completed, the changes in the patient's physiological indicators and infusion data are collected. Based on the changes in the patient's physiological indicators and infusion data, the infusion effect is evaluated. The infusion effect data and the corresponding decision data are used together to optimize and update the intelligent matching algorithm and dynamic adjustment logic.

Images